fluent uses only 1 core

[ilfrau]

hello to all
I am using fluent for the first time, in the internal version of ansys 13, for a university relationship aimed at learning the basics of the program (as no course teaches you to use it). The speech is long... .
In any case yesterday I launched a simulation and, by chance, I also had the task manager open and I saw that the use of the cpu went from 25-27% and just.... which left me a little puzzled because, from what I read, the fluent is cpu intensive. . .
my configuration is:
ii x4 955
4 giga di ram
os w7 64 bit
The rest I think doesn't matter

the ram me occupies it to 80% with even higher tips, but the cpu no...look a little 25% is 1/4 of 100 and I have 4 cores...so I thought the program used only 1
searching on the net I found discussions that confirmed that it can happen

How do I do that? parallel? but isn't it just in more pcs? How do you do that?
Thank you.

 

[sail]

hello to all
I am using fluent for the first time, in the internal version of ansys 13, for a university relationship aimed at learning the basics of the program (as no course teaches you to use it). The speech is long... .
In any case yesterday I launched a simulation and, by chance, I also had the task manager open and I saw that the use of the cpu went from 25-27% and just.... which left me a little puzzled because, from what I read, the fluent is cpu intensive. . .
my configuration is:
ii x4 955
4 giga di ram
os w7 64 bit
The rest I think doesn't matter

the ram me occupies it to 80% with even higher tips, but the cpu no...look a little 25% is 1/4 of 100 and I have 4 cores...so I thought the program used only 1
searching on the net I found discussions that confirmed that it can happen

How do I do that? parallel? but isn't it just in more pcs? How do you do that?
Thank you.

I press that they are on linux and launch it using the "calx" system (no workbench), so it could be a little different. ...

the launch pattern, where you have to specify whether the problem is 3d or 2d, and the solutor precision has at the bottom right of the options to choose whether to turn it in serial or parallel mode. if you select the parallel mode on the local machine will ask you to specify the type of interconnection (default will go well), the type of message passing interface (I use openmpi, but there are other...), the distribution of memory (you want shared memory) and the method to throw the solutors on the various porcessors (ssh or rsh).

try a little, if you need place some screenshots.

 

[ilfrau]

Thanks a lot
I did as you told me, with default settings, and the cpu went 100%. I've finished my simulation in an hour!

I can also ask for another thing though....
as I said use ansys 13, in the workbench there is a meshing environment that is called "meshing"...I don't know what algorithms you use since, always in the package of ansys, there are other more dated meshators like turbogrid or icem (never used....as they are at first weapons)
In any case even in that environment when I launch the mesh command, it only uses 25% cpu. but here I can't do as in fluent.. .
in the options menu there is a voice saying "number of cpus"; by default it is set 0.... but also after having put it to 4 (as shown in the attachment) nothing changes. .

 

[sail]

Thanks a lot
I did as you told me, with default settings, and the cpu went 100%. I've finished my simulation in an hour!

I can also ask for another thing though....
as I said use ansys 13, in the workbench there is a meshing environment that is called "meshing"...I don't know what algorithms you use since, always in the package of ansys, there are other more dated meshators like turbogrid or icem (never used....as they are at first weapons)
In any case even in that environment when I launch the mesh command, it only uses 25% cpu. but here I can't do as in fluent.. .
in the options menu there is a voice saying "number of cpus"; by default it is set 0.... but also after having put it to 4 (as shown in the attachment) nothing changes. .

It's normal. fluent uses multiple processors by subdividing the domain into small parts and assigning each one to a processor and using the mpi to communicate what happens to the interface.

Meshatura is a phase that unfortunately does not take advantage of multitreading, so you will always and only use a core.

 

[Flac]

Hello everyone, I have the same problem with fluent 6 on linux (ubuntu). I have a quad core processor, but unfortunately, looking at the system monitor during simulations, I realized that I can't fully exploit the cpu, in this case only one core is exploited and not 100%. I tried to put in place the suggestion proposed in the previous thread, but I couldn't find the suggested setting.
Thank you in advance for the answer
cordially
flax